Development of upper-layer protocols with S2CR acoustic modems emulator

Transcription

1 Development of upper-layer protocols with S2CR acoustic modems emulator Oleksiy Kebkal EvoLogics GmbH Berlin, Germany Konstantin Kebkal EvoLogics GmbH Berlin, Germany Maksym Komar EvoLogics GmbH Berlin, Germany Abstract Modern underwater acoustic modems are sophisticated devices, designed to help for solving challenging practical tasks, like monitoring the underwater infrastructure in offshore industry, positioning and data exchange during cooperative missions of underwater vehicles, precise positioning of distributed underwater nodes. Most of these applications are unique, and each requires development of complex software with implemented custom protocols for data exchange between the nodes of underwater acoustic sensor network. The software has to meet strict requirements to its long-term reliability and must undergo extensive testing. As deployment, maintenance and recovery costs of underwater systems are notably high, to keep the development costs within reasonable limits, development and implementation of the software should be preferably done by means of a highquality emulation/simulation environment. The paper describes a real-time emulator of EvoLogics underwater acoustic modems. The emulator provides a fully-featured emulation of the data-link layer and includes a simplified simulator of the physical layer that accounts for signal propagation delays, multipath propagation, data packet collisions, packet synchronization errors and bit errors with a user-defined bit error rate. The emulator substitutes acoustic modem hardware when running testbed scenarios during development and testing of the upper layer protocols and applications. The emulator provides fully featured support of its cross-layer synchronization mechanism, allowing to develop positioning protocols or hybrid protocols combining communication with positioning on the upper layer protocol. I. INTRODUCTION Modern underwater acoustic modems are sophisticated devices that support not only point-to-point data exchange between two systems, but also enable a complex integration of underwater acoustic sensor networks. These acoustic devices must solve challenging practical tasks, like monitoring underwater infrastructure of offshore industry, positioning and data exchange during cooperative missions of underwater vehicles, precise positioning of the nodes of distributed underwater systems etc. Most of these applications are unique, and each requires development of complex software implementing custom protocols for data exchange between the nodes of an underwater acoustic sensor network. The software has to meet strict requirements to its long-term reliability and must undergo extensive testing. As deployment, maintenance and recovery costs of underwater systems are notably high, to keep the development costs within reasonable limits, development and implementation of the software must be performed in a highquality simulation environment. Existing network simulators, for example, the NS2- Miracle[1], are highly useful tools for research and development, but a network simulator alone is not sufficient for many aspects of software development including algorithm design, simulation, verification and testing of the target system. An essential step forward was made with the recent releases of network simulator extensions, such as the SUNSET[2] and the DESERT[3], which enable interfacing real acoustic modems into the testbed framework, substituting the simulations of the physical and data-link layers. Nevertheless, this approach has its shortcomings. Using expensive real modem hardware for simulation-related purposes elevates development costs. On the other hand, the NS2- Miracle itself does not implement physical and data-link layer protocols of commercially available modems, Therefore, the difference between the real hardware and its implementation in the simulator, the difference in the interfaces in the simulator and the hardware and other factors, specific for the simulator, cause an essential difference in implementation of the same upper layer protocol for the simulation environment and for the real hardware. In this paper, we describe a real-time emulator of EvoLogics S2CR series underwater acoustic modems. The emulator provides a fully-featured emulation of the data-link layer and includes a simplified simulator of the physical layer that accounts for signal propagation delays, multipath propagation, data packet collisions, packet synchronization errors and bit errors with a user-defined bit error rate. This emulator can substitute acoustic modem hardware when running testbed scenarios on a network simulator. In addition, it can serve as a standalone solution for development and verification of custom network protocols. The emulator fully supports the cross-layer synchronization mechanisms implemented in S2C modems, and thus allows to develop positioning protocols and hybrid protocols that combine communication with positioning on a upper layer protocol. In the following sections, we will also elaborate on applying the code developed with the emulator to real acoustic modems. The rest of the paper is organized as follows. Section II describes the S2C protocol stack, implemented in the Evo-

2 Logics acoustic modems. Section III presents the concept of the acoustic modems emulator and explains the purposes of its development. Section IV addresses the practical issues of the whole development process, from modeling to sea trials. Section V describes the practical experience of using the emulator. Section VI outlines the plans for the emulator s future development, the final section VII sums up the paper. II. INTRODUCTION TO S2C PROTOCOL STACK The S2CR series underwater acoustic modems, manufactured by EvoLogics GmbH, comprise the following components: transducer with the transmit/receive amplifier a digital stack optional USBL antenna optional Wake Up module The physical properties of the transducer type define the beam pattern of the acoustic modem and the frequency range. Both the properties define the range of the applications, where the particular model of the acoustic modem can be used. The acoustic modems comprise also the matched transmit/receive amplifier, optimized according to the characteristics of the particular transducer type. The ultra-short baseline antenna (USBL-antenna) is an optional module comprising a grid of 5 receivers integrated together with the transducer into one housing with the receive amplifiers for each channel of the grid. The Wake Up Module is an optional integrated electronic unit that can turn parts of the device off to save power. The Wake Up Module helps optimizing power consumption for battery-powered deployments by checking for incoming acoustic signals or incoming data on the host interface and turning rest of the device on only when such a signal is detected. Once the device completes receiving or transmitting data, everything but the Wake Up Module is switched off. The digital stack consists of the ADC, DAC, DSP and FPGA that implement the physical layer protocol, further referred to as S2CPhy, and a host processor, implementing the D-MAC data-link layer protocol. We will describe both protocols in the sections below. A. S2C physical layer protocol The physical layer implements the patented S2C (Sweep Spread Carrier) signal modulation technique. S2C is based on the assumptions that, for an underwater acoustic channel, a received acoustic signal is well described by a sum of multipath components with random amplitudes and phases, and that the multipath intensity profile is discrete. In these conditions, one of the most promising approaches to digital communication is the spread spectrum modulation. A spread spectrum signal is characterized by a short autocorrelation response. The broader the spectrum spread, the shorter its response. After passing the underwater acoustic channel, the received signal is a sum of multipath components and, after matched filtering, it can be presented as a series of time-shifted correlation responses. For a signal with broad spectrum spread, these responses can be isolated, eliminating the signal distortion associated with multipath propagation. In particular, the phase distortions of an isolated multipath component can be significantly reduced. S2C method utilizes broad-spectrum signals with continuously varying frequency. These signals are well known in radio location, where they are used for increasing the probability of successful signal detection. It is known, that if the such a signal has a large base, at least 95 percent of its energy is concentrated in the given frequency band, and its auto-correlation response is short. S2C technique utilizes the method of continuous carrier frequency variation, where the information is coded by discrete manipulation of one or several carrier parameters. In contrary to other common methods of digital underwater acoustic communication, an S2C signal is characterized by two levels of modulation: first the internal modulation for continuous variation of the carrier frequency (analog modulation), second the external modulation for coding the information within the signal (discrete manipulation). With the frequency band ranging from one to tens of khz, the signal can be just hundreds of microseconds long, and the transmission speed can reach tens of kbits per second. Implementation and application of the S2C method demonstrate one of its key advantages: with continuous carrier modulation instantaneous frequencies of the incoming multipath components are shifted away from the instantaneous frequencies of previously received multipath components. Therefore, the band-selective distortions of the received signal are weakened. This leads to increased bitrate, reliability and efficiency of the data transmission. The key concepts of the S2C method are implemented on the physical layer of the S2C protocol stack, the S2CPhy. This protocol is implemented in the DSP and the FPGA of the digital stack and performs the following tasks: 1) estimation of the parameters of the underwater acoustic channel: obtaining the multipath intensity profile (evaluating the intensities and the propagation delays of multipath components); selection of the most intensive component for receiver synchronization; determining the propagation delay and the Doppler shift of the most intensive component; 2) packet and symbol synchronization evaluating the beginning and the end of a data packet; evaluating the reception time of every symbol in the series (accounting for the Doppler shift); 3) modulation: transforming the sequence of bytes into a sequence of words according to the manipulation rate, in particular, transforming into a sequence of bit pairs; defining the signal envelope according to the modulation type selected;

3 defining the signal envelope according to the carrier sweep; 4) demodulation: synchronized processing of the received signals with the in-phase and squared components of the reference signal; evaluation of the complex envelope of the received signal evaluation of the discrete signal value; determination of the bit pair that corresponds to a discrete phase value of the received signal. 5) positioning: evaluation of the pulse response of the channel to select a multipath component that corresponds to the shortest propagation path; determination of the time difference between the acoustic front incidence on the elements of the USBL grid. B. D-MAC data-link layer protocol Design of an improved data-link layer protocol called D- MAC is based on the recently developed data delivery algorithms of acoustic modems that implement the S2C Technology. According to the new data-link layer protocol, two different types of data are supported, namely burst data and instant messages [4]. Burst Data: Establishing a connection for burst data delivery requires an estimation of the channel s parameters. As described in [5], the delivery algorithm optimizes the channel s utilization efficiency and adapts the bitrate to the highest possible value for a particular underwater acoustic channel. All data received from the user is buffered and then dynamically split into smaller packets according to channel parameters. The receiver side assembles the split data together and sends it to the remote user in the original format. Instant Messages: Establishing a connection is not required for delivering instant messages. A fixed bitrate (relatively low and acceptable for a wide range of acoustic channel parameters [4]) is used for delivering short instant messages, as the algorithm minimizes delivery time of such messages. Delivery of instant messages doesn t interrupt the ongoing burst data transmission, since instant messages are delivered as an extension of service messages. An instant message can be O(10 2 ) bits long. On one hand, a low bit error rate implies a low transmission rate so the instant messages tend to be long to transmit. On the other hand, duration of a message must be less than the channel coherence time [6]. Thus, the bitrate must be high enough to make sure that the message duration meets the time constraints, forced by the channel coherence time. A physical layer protocol, based on the S2C technology [7] provides reliable transmission of instant messages with a 1 kbps bitrate. This is valid even for channels with highly dynamical parameters like confined water bodies with moving network nodes. TABLE I CLASSIFICATION OF INSTANT MESSAGES Instant Message Asynchronous Synchronous Robust Unicast Broadcast Robust Unicast Broadcast On the data-link layer, message delivery time can be made as short as possible by transmitting the message with a fixed low bitrate without handshaking and adaptation of the communication system to channel parameters. Instant messages can be classified according to message addressing type, acknowledgment and synchronization requirements. Table I displays a classification of instant messages. Asynchronous instant messages delivery is based on an ALOHA-like scheme when there is no ongoing burst data exchange between the nodes of an acoustic network. Asynchronous instant messages can also be delivered as parts of service messages of the burst data delivery protocol. Media access control for synchronous instant messages transmission must be implemented by upper level protocols. To fulfill this task, the D-MAC protocol implements a custom interface for synchronization with the physical layer, allowing the upper level protocols to point out the time for transmission of a synchronous instant message and also obtain the message arrival time. Synchronous instant messages cannot be transmitted during burst data exchange. III. S2C MODEM EMULATION CONCEPT The major purpose of the modem emulator is to minimize development costs of upper layer protocols and to simplify the integration of acoustic modems into underwater infrastructure. The main consideration for modem emulator design was that an application, developed with the emulator, must work with real acoustic modems without any code modifications. This defines the following requirements the modem emulator must meet: real-time emulation of a large number of network nodes; same source code for both the emulator and the real modem firmware; equal command set for both the emulator and the modem; remote emulator access via Internet. The requirement for real-time emulation directly derives from the main purpose of the emulator. Time diagrams of both the emulator and the modem must match to ensure the protocols and upper layer applications operate identically on the emulator and the modem. One must take into consideration, that upper layer protocols can utilize both the data exchange functionality of the modems and the distance measurement functionality to solve positioning tasks either in parallel with the data exchange or in its absence. The modem s firmware is constantly evolving, since, as the modems find new applications, they support integration with a growing range of external sensors and other devices. The datalink layer is the one that gets the most changes, and it is the one most visible to upper layer protocols. If the emulator source

4 code significantly differs from the firmware, every firmware update would, first, require a modification of the emulator code and, second, would demand a wide range of tests performed to verify the identity of the emulator and the firmware. A more effective approach, as it turns out, is to launch the code, developed for the modem firmware, on another platform, namely the x86. The platform-dependent part of the code was contained within a compact driver, implemented in two versions the one for the modem firmware and the other for the x86 emulation. Utilizing the same source code for both the firmware and the emulator also guarantees the identity of the command sets that control them, making them indistinguishable for upper layer protocols. A multitude of instances of acoustic modem emulator comprising an underwater acoustic network can be configured and launched on the manufacturer s server. The user is granted remote access to this acoustic network. Each modem can be accessed via TCP/IP socket. This approach ensures prompt updates of the code and makes the usage of the emulator platform-independent, as the user does not need additional equipment to install and run the emulator. According to the design objectives, the components of the modem emulator are the following: the data-link layer the core module that controls the data exchange between the data-link layer and the physical layer the physical layer simulator the acoustic channel simulator As mentioned above, the emulator and the real acoustic modems use the same data-link layer source code, compiled for the target platform. The platforms supported are the ARM, x86, x This approach saves time and effort spent on support and development of the emulator, ensures full compatibility of the data-link layer protocols of the emulator and the modems, reducing development time for system integration of the modems or, for R&D purposes, shortens the path from simulations to final trials. Cross-layer communication between the data-link and the physical layers is provided by a Linux-core driver that implements a platform-dependent code, specific for the acoustic modem hardware. This code provides low-level access to the data exchange interfaces of the data-link and the physical layers. Within the emulator, this driver redirects the datalink layer s requests back to the user-space of the physical layer simulator and back from the simulator to the data-link emulator. This data-link layer implementation, separated between a platform-dependent core driver and the main POSIXcompliant code, enables using the same code base for the modem emulator and the real modem. The physical layer simulator imitates it by replying to the data-link layer s requests according to the cross-layer data exchange specification. The simulator s parameters are its three-dimensional coordinates and the acoustic channel s bit error rate. The simulator transfers the data to be transmitted to the dispatcher that enables the data exchange between modems. The simulator includes the time-stamp of transmission start and its own coordinates within the data. Having received data packets from the dispatcher, the physical layer simulator imitates a propagation delay by holding the packet for a timeout that corresponds to the signal propagation time between the signal source and the receiver, and detects collisions, dropping the collided packets. To calculate the propagation time, it is possible to introduce a sound speed profile for testing positioning protocols. The last component, the acoustic channel simulator, is effectively the dispatcher of the data-link layer packets. The main function of the component is to receive the packets from a physical layer simulator and forward them to the other physical layer simulators, connected to the dispatcher. IV. FROM EMULATION TO SEA TRIALS In most publications on communication protocols for underwater acoustic sensor networks, known to the authors of the paper, experimental results were obtained by simulations with the NS2 and NS3 the network simulators, well-known in the academic community. Experimental results from realworld trials, on a lake or at the sea, is rather exceptional. One of the crucial obstacles for real-world experiments is the high cost of both the underwater equipment and the vessel time at sea. Nevertheless, there is an another equally important factor, namely the essential difference between protocol implementation for simulation and for real-world experiment purposes. A big step forward from modeling towards experimental studies was made with the recent releases of SUNSET and DESERT frameworks, both based on NS2-Miracle[1], capable of working with real modem hardware in testbed scenarios. A common source code base can now implement upper layer protocols for simulations and for testbed scenarios with real hardware. The only link missing was a fully-featured acoustic modem emulator, eliminating the need for real modems during the R&D and testing, as well as extending the number of possible test scenarios, hardly manageable even with real modems at the development stage. In particular: the emulator allows arbitrary propagation delays between network nodes, while deploying real modems on essential distances is too much effort for development purposes; the emulator supports real-time testing of multiple underwater acoustic network nodes at once, while several dozens of modems, batteries, buoys, anchor chains and other accessories are an unaffordable luxury for a network protocol developer; the emulator fully supports the S2C modems cross-layer synchronization mechanisms, essential for implementing

5 positioning protocols that consume too much time, cost and effort for real-life R&D testing; the physical layer simulator supports collision detection and user-defined demodulation and synchronization error rates, allowing to test applications and upper layer protocols in different operating conditions and debug the code to improve system stability without involving expensive underwater infrastructure. The next step paving the way from modeling to open-sea trials was the launch of EvoLogics White Line Science Edition underwater acoustic modems that provide the user with a firmware sandbox for launching custom applications or upper layer protocols. The sandbox allows to run tcl/expect scripts for quick development of test scenarios or applications, as well as to launch C/C++ applications or the SUNSET or DESERT frameworks. V. USAGE EXPERIENCE The emulator of the S2CR underwater acoustics modems series underwent beta-testing last year. Our academic partners were granted access to the emulator to develop and debug upper layer protocols and get ready for joint trials. Furthermore, commercial customers were granted emulator access to simplify system integration of S2C modems they purchased: the customers were able to use the emulator and get accustomed with the modems before the actual hardware delivery. This August, the authors of this paper conducted joint trials with the SIGNET group from the University of Padova (Italy). The purpose of the trials was to test the SIGNET group s SUN[8] protocol a dynamic source routing protocol for underwater acoustic sensor networks. Our colleagues from the University of Padova already had strong experience with the S2C emulator. Via remote access, they had used it to debug and test the SUN protocol within the framework of a researchoriented NS-Miracle network simulator. Because of that experience, preparing for the field tests, namely the transition from the S2C emulator to real S2C hardware, went easily and took just a few days. The trials involved the White Line Science Edition modems (the S2CR WiSE), and were held at the Werbellinsee lake near Berlin, Germany. The test results were presented in [8]. During their research and development process, our colleagues from the University of Padova found several uses for the emulator and provided the following feedback [9]: the emulator is a learning tool, as it allows one to get confident with the AT commands that control S2C modems; it is very useful for designing and refining sophisticated interactions with the modem, for example, the emulator can be used to test and improve network protocols and applications that involve several message exchanges with the modem on the basis of, e.g., a suitably designed finite state machine; the emulator allows to plan and refine experiments before conducting them on real hardware. Based on this experience, the following benefits of working with the emulator can be noted: the emulator enables the researchers and developers to test their network protocols and/or application solutions without the modem hardware directly connected to their computers, moreover, it allows to use several different (virtual) modems at once and provides the freedom of working over remote access; Using the emulator saves time during code debugging and refinement. It is quicker and much easier to use one s own computer to modify the code and test it than to use the actual modem for this purpose; Furthermore, solutions that were designed and tested with the emulator are easily exportable to the actual modem hardware. VI. FUTURE DEVELOPMENT The underwater acoustic modem emulator is constantly evolving, since the growing experience of its application inspires new ideas for further developments of the original concept. Having used the emulator, our partners contributed with some suggestions that helped devise the development plan introducing the following features: user-friendly customization of the emulator settings using web-based configuration utilities; a user-friendly topology constructor to set up the actual network topology; a log visualization tool, allowing to view and analyze the events log (e.g., packet flows, packets disregarded for errors and/or collisions); supporting mobile network nodes with custom userdefined mobility models; user-defined parametrization of the synchronization and demodulation error as a function of distance between nodes; possibility to run the emulator as a virtual machine on the user s computer. VII. CONCLUSIONS The emulation approach described above was developed to open more opportunities for researchers and developers of upper layer protocols and other applications that use underwater acoustic modems. The ability to test and debug an application with a modem emulator allows to significantly reduce development costs and increase the reliability of the solutions developed, as open-sea trials require too much time, cost and effort. The emulator usage experience, both during internal testing and within a partner cooperation, proved it a valuable solution. The transition from using the emulator to deploying real acoustic modems took from days to a week, depending on the complexity of the experiment and the need for personnel training.

6 The emulator can be used either as a standalone solution for research and comparison of different approaches and upper layer protocols, or as an additional tool to speed up the development process. ACKNOWLEDGMENTS The authors would like to thank the SIGNET group of the University of Padova (Italy) and Dr. Riccardo Masiero for the interesting discussions and for the feedback about emulator usage. Thanks also go out to Ms. Mariia Pleskach for valuable remarks during preparation of the paper. REFERENCES [1] (2012, Aug.) The network simulator - ns-2. [Online]. Available: User Information [2] C. Petrioli, R. Petroccia, J. Shusta, and L. Freitag, From underwater simulation to at-sea testing using the ns-2 network simulator, in OCEANS, 2011 IEEE - Spain, june 2011, pp [3] R. Masiero, S. Azad, F. Favaro, M. Petrani, G. Toso, F. Guerra, P. Casari, and M. Zorzi, Desert underwater: An ns-miracle-based framework to design, simulate, emulate and realize test-beds for underwater network protocols, in OCEANS, Yeosu, may 2012, pp [4] O. Kebkal, M. Komar, K. Kebkal, and R. Bannasch, D-mac: Media access control architecture for underwater acoustic sensor networks, in OCEANS, 2011 IEEE - Spain, june 2011, pp [5] O. Kebkal, On the use of interwoven order of oncoming packets for reliable underwater acoustic data transfer, in OCEANS EUROPE, may 2009, pp [6] B. Sklar, Digital communications: fundamentals and applications, ser. Prentice Hall Communications Engineering and Emerging Technologies Series. Prentice-Hall PTR, [7] K. G. Kebkal and R. Bannasch, Sweep-spread carrier for underwater communication over acoustic channels with strong multipath propagation, The Journal of the Acoustical Society of America, vol. 112, no. 5, pp , [8] G. Toso, R. Masiero, P. Casari, O. Kebkal, M. Komar, and M. Zorzi, Field experiments for dynamic source routing: S2c evologics modems run the sun protocol using the desert underwater libraries, in OCEANS, Hampton Roads, in press. [9] R. Masiero, personal communication, 2012.